- From: Bijan Parsia <bparsia@cs.man.ac.uk>
- Date: Fri, 11 Jan 2008 13:28:46 +0000
- To: Jeremy Carroll <jjc@hpl.hp.com>
- Cc: Owl Dev <public-owl-dev@w3.org>
On 11 Jan 2008, at 10:56, Jeremy Carroll wrote: [snip] > A) classes as instances > ======================= > A class in one view of an ontology may be an individual in another. > e.g. > the class of Ford Motor Vehicle Models, may include the individual > Ford Cortina; > the class of Ford Cortina may include a particular car that I once > owned (actually an untrue statement!). > > It can be argued that this is a modelling error. However, if so, it > is a popular one. It also comes up in ontology integration and reconciliation (i.e., you modeled something as a class, I modeled the same thing as an instance, I'd like to merge our ontologies *first* and work with the merged one where it's used both as a class and as an instance so that I can better see the effects of changes as I work to reconcile the points of view) or in some modeling patterns. (Arguably, those patterns are work arounds or "hacks" in many cases, but it's important to provide such escape hatches.) [snip] > B) RDF legacy > ============= > dc:creator is used as both, and its definition seems to expect it > to be used as both. There are perhaps some other properties from > the RDF world that are important and behave similarly. (e.g. > rdf:_1, rdf:_2) > > Technically, but of no great interest really, RDF legacy also > includes some 'punning' between individuals and classes, because > RDF includes metamodelling such as: > owl:inverseOf rdf:type rdf:Property. > > > C) lists > ======== > Evren provided the example of rdf:first (or eg:first in a shadow > vocabulary), in a list. In a list of literals, rdf:first acts as a > data property, in a list of things, rdf:first acts as an object > property. > > Bijan motivates some of the design decisions in OWL 1.1 as trying > to accomodate more RDF, so that arguments B and C have merit. Also, > because of the popularity of the modelling technique of classes as > instances, A has merit. > [[Bijan, have I correctly understood your position?]] All other things being equal, yes. The integration argument works for *all* punning, so my motivation isn't "merely" technical (to accommodate more RDF graphs as an end in itself), but user driven (people do need to have ontologies which are in a non-separated state, if sometimes temporarily). If I were convinced that there were cases where it was super duper unlikely that users wouldn't care (e.g., I could be convinced that datatypes should be separated) then if there were some technical reason to forbid that punning I could be convinced. I'd be surprised if the loss of regularity would have any interesting benefits. But yes, if: http://www.w3.org/2002/07/owl.rdf and http://www.w3.org/2000/01/rdf-schema.owl were entirely OWL illegal, I wouldn't lose any sleep. OTOH, if they can be accommodated in a simple, straightforward way, I can live with that and would support it (though without too much enthusiasm :)). I would, however, vigorously denounce them and their ilk as lame, pointless, misleading, and so on in all fora :) (On technical grounds, of course!) Cheers, Bijan.
Received on Friday, 11 January 2008 13:26:54 UTC